Image courtesy https://www.flickr.com/photos/thibaultj/
Completing a PhD involves more than hard work; it exemplifies the idea that the whole is greater than the sum of its parts. While this is true, it is important to recognize that each component significantly contributes to the outcome. Researchers now have access to various tools, from PCR machines to sophisticated microscopes, which enable rapid groundbreaking discoveries. However, as data acquisition has become easier, the scale of data collection has also increased exponentially. This shift has led to an expectation for PhD students to tackle more complex scientific questions, thereby increasing the workload necessary to complete projects and publish results.
Artificial intelligence (AI) tools can help graduate students tackle complex biological questions more efficiently by completing tasks in seconds that used to take hours. One of the most popular and accessible AI tools is known as ChatGPT, which has been my (and others) introduction to the power of conversational Large Language Models. In brief, ChatGPT is a conversational AI that can answer questions and have discussions that feel natural to humans.
We have previously covered AI on Signals in two posts: science communication and impacts of AI on health care/biomanufacturing. Now, I intend to focus on how graduate students can harness the utility of AI tools like ChatGPT. While a premium version is available, everything I will discuss was done using the unpaid version of ChatGPT to ensure accessibility.
Data analysis and visualization
In my research, I conduct proteomic experiments to explore the relationships among proteins in human cells. The human genome can encode approximately 20,000 proteins, leading to large datasets even when only a fraction is detected. These datasets are often outputted as large spreadsheets, which require Excel scripts for effective organization and analysis. I was relatively inexperienced with this, so I turned to ChatGPT for assistance.
Describing this scenario, in plain text, helped me generate formulas to organize my data, such as filtering proteins based on abundance and p-value cut-offs. This helped me efficiently make sense of my data and concentrate on the most relevant proteins for further analysis. In parallel, I was also ensuring that I was employing appropriate statistical methodologies and comprehending their strengths and weaknesses.
With my studies nearing an end, I plan to use ChatGPT to help me decide what method will present my research data most effectively. While it is not yet a feature to produce visualizations (maybe soon!), ChatGPT can use descriptions of data to suggest an appropriate visualization and even guide me on how to produce them. While these suggestions are only scratching the surface of what is possible, they represent simple starting points for integrating AI into your data analysis pipeline.
Reading manuscripts
The research landscape is rapidly expanding, with several new papers published weekly across thousands of specialized journals. For instance, I have alerts for 20 terms related to my research, which means approximately 100 papers land in my inbox weekly. It is not feasible to analyze every single one extensively. This is another area where ChatGPT has proven to be incredibly valuable in my PhD research.
When I encounter a manuscript with complex methodologies or theories that are difficult to grasp, I turn to ChatGPT as a sounding board to help clarify and break down the information. Instead of spending hours searching through Google and carefully reading each manuscript, ChatGPT acts like a research assistant, helping me efficiently make sense of the latest findings and stay updated with the ever-changing scientific landscape.
Enhancing writing
Writing award applications, manuscripts, progress reports and presentations requires clearly communicating complex ideas. I have learned to use ChatGPT to help me, making me a better writer. One simple application is using ChatGPT to help brainstorm titles for my presentations and reports. While titles can be an afterthought, they set the tone and need to convey the proper message to orient readers and audiences. To craft powerful titles, I prompt ChatGPT to provide me with 20 examples and then combine aspects of a few of them to ensure they are concise and meaningful.
ChatGPT can also serve as a real-time editor when drafting a document, giving me feedback on any inconsistencies and areas needing correction. This is crucial because each document has a different purpose, ranging from general science communication to detailing research findings in a manuscript. Therefore, each requires a separate assessment for the proper clarity/conciseness, tone, flow and overall impact, which ChatGPT excels at.
Another area where ChatGPT shines is career planning. Being surrounded by academics, receiving feedback tailored to drafting a resume and/or cover letter relevant to industry roles is often difficult. I have used ChatGPT to update my resume and ensure I am always highlighting the relevant skills I have attained through my experiences. Currently, I am taking this one step further and practicing writing cover letters for roles I plan to pursue upon graduation.
Ethical considerations
It is important to emphasize that while ChatGPT is a powerful tool, it should be used as an aid rather than a replacement for original research and writing. Graduate students must take responsibility for their work, ensuring it reflects their intellectual contributions. Accuracy is also a key concern, especially in scientific endeavours and using any online resource, therefore it is essential to fact-check information from ChatGPT. While I have found its responses to be quite accurate, it makes errors, so I always cross-reference the information I receive.
Additionally, the use of ChatGPT may be subject to rules from departments or journals, so it is important to be aware of, and comply with, any guidelines. Privacy is another consideration, as all data generated during interactions with ChatGPT are owned by OpenAI, the company behind it.
Looking ahead
I have purposely not mentioned many other features of ChatGPT since this blog is meant to be simple and focus mostly on stimulating interest. I have also avoided discussing other alternatives to ChatGPT, but they do exist. A final tip is to always check for free premium versions through institutional access for ChatGPT and alternatives. They are more powerful and faster, so take advantage if available.
What are your thoughts, readers? I would love to hear your suggestions on how to further introduce graduate students to the power of AI. Please share your comments below.
Francesco Zangari
Comments